Professional Documents
Culture Documents
General Overview of Spatial Impression...
General Overview of Spatial Impression...
Lexicon,Bedford,MA,USA
This paper provides an overview of the current knowledge of the problems of spatial impression, envelopment, and
localization in enclosed spaces. The emphasis will be on small spaces, but the approach draws heavily on concert
acoustic studies. The literature on this subject has been contradictory and controversial. In the last few years a consen-
sus has been developing that envelopment is primarily determined by lateral reflected energy arriving at least 80 ms af-
ter the direct sound. Localization is traditionally linked to the apparent source width (ASW), which has been associated
with the interaural correlation (IACC) of the first 80 ms of the impulse response of a space. Although ASW and IACC
may be useful in concert hall measurement, they do not work well in small spaces. This paper will present two new
measures: the diffuse field transfer function (DFT) and the average interaural time delay (AITD). The DFT is a mea-
sure of envelopment, and is useful both in small and large rooms. The AITD is a measure for "externalization," a sonic
property unique to small rooms. A closely related measure, the net interaural time delay (NITD), is useful in under-
standing localization in small spaces.
In the frequency band between 300Hz and (Ando and Morimoto use the interaurai cross
3000Hz envelopment is determined by a correlation (IACC) as one of their measures of
combination of fluctuations in the IID and the the spatial properties of concert hail sound.
lTD. This fact was discovered independently by Morimoto uses the IACC of the first 80ms of a
the author and Blauert. Blauert notes in [7] and binaural impulse response as a measure.
[8] that fluctuations in IID have slightly different Beranek and Hidaka call this measure (IACCe).
perceptual properties than fluctuations in lTD. For reasons described extensively in [44], the
He concludes that different brain structures are author finds the measure often misleading,
involved in the two perceptions. Envelopment at particularly in small rooms. However the IACCe
frequencies below 300Hz is determined almost does describe the high frequency angular
entirely by fluctuations in the lTD. The dependence of envelopment quite well.)
important virtue of the framework presented in
[44] for the perception of envelopment is that the Using interaural fluctuations it can also be
concept of interaural fluctuations takes the study shown that if there are two sound sources in an
of envelopment out of the realm of anechoic space, and uncorrelated band-limited
psychoacoustics and into the realm of physics, noise is played through each source, the sense of
We can model the mechanisms that create these envelopment will be maximum when both
fluctuations, and we can measure them. sources are at the optimal angle for the frequency
band in question. For frequencies below 700Hz
Once we have moved the problem into the realm the optimal angle is 90 degrees - the sources
of physics, we can see that the problem of must be at the sides of the listener. As the
envelopment has two parts - the nature of the frequency rises, the optimum position moves
envelopment receiver, and the nature of the toward the medial plane.
sound field surrounding the receiver.
The angular dependence of envelopment at high
The receiver for envelopment is the human head, frequencies has important consequences for
the outer and inner ears, and the brain. The sound reproduction systems. A few years ago
soundfield is the combination of direct and the author was playing a stereo tape of applause
reflected energy at the listening position. The through a standard two channel system in a
head/ear system is the antenna for the perception relatively dead room. He was very surprised to
of envelopment. Like other antenna, this system hear the applause coming from ail around him,
has directional properties. We can study the even though the loudspeakers were clearly in
directional dependence of the head/ear system front. Adding a bandpass filter quickly showed
for single reflections by convolving test signals that the perception was a simple result of the
with published Head Related Transfer Functions angular dependence of high frequency
(HRTFs) such as the ones put on the web by Bill envelopment. The 1700Hz band produced a
Gardner. After the convolution we can detect surrounding perception, even though the
and measure the interaural fluctuations that result loudspeakers were at +-30 degrees. Because the
using the software detectors that will be applause had significant energy in this band, the
described later in this paper, result was enveloping. We conclude that
stereophonic reproduction works in part because
This work is yet to be done. However several envelopment in a particular frequency band can
years ago the author did models of the high give an impression of envelopment in all bands,
frequency behavior of interaural fluctuations and even with a relatively narrow front
using a simple head model. We found that the loudspeaker pair, envelopment at some
resulting fluctuations depend on the azimuth of frequencies will be high.
the reflection. For 1000 Hz the optimum angle
lies in a cone centered on a line drawn through Morimoto's front/back ratio becomes important
the listener's head. This cone makes an angle of when the reflected sound contains significant
about 60 degrees from the front of the head. At energy above 3000Hz. In the author's
about 1700Hz the cone includes the standard experience with concert hails and opera houses it
loudspeaker positions at +- 30 degrees from the is rare that there is significant energy at those
AES
15th
INTERNATIONAL
CONFERENCE 139
DAVID GRIESINGER
frequencies coming from the rear. Sound from foreground and a background stream. Where
the rear of a hall is often spatially diffused, and this separation is possible, it is the spatial
has been reflected off many surfaces. Typically properties of the background stream -
each surface takes a toll on the high frequency specifically the amount of interaural fluctuations
content. In halls where an electronic - that determine the envelopment we perceive.
enhancement system has been installed it is In a small room there is almost never sufficient
possible to experiment with the frequency late reflected energy to contribute to the
spectrum of the reflected energy, and invariably background perception. The late energy must be
the sound is better if frequencies above 3000I-Iz supplied by reverberation in the original
are attenuated. Thus the author is not convinced recording. To predict the degree of envelopment
that the front/back ratio is an important measure we perceive we must be able to predict the
for concerthall design, strength of the interaural fluctuationsduring the
reverberant segments of the recording.
However Morimoto has shown conclusively that
the front/back ratio is important when listening Recording engineers know that for best results
with loudspeakers, and the author can confirm the reverberation in a two channel recording
the observation. This issue becomes quite should be uncorrelated - completely different in
important in the design of surround sound the left and right channels. It is the job of the
playback systems. Several of the best sound loudspeaker/room system to cause the listener to
mixers for surround sound have noted that they have adequate interaural fluctuations when this
prefer the rear loudspeakers to be placed +- 150 condition occurs. The loudspeaker/room system
degrees from the front, and not the more typical is acting as a transfer system, transferring the
+-110 degrees. At 150 degrees these decorrelation in the recording to the listener's
loudspeakers are capable of producing the ears. We need a measure for how effectively this
frequency notches in the HRTFs that indicate transfer works.
rear sound, and speakers at 110 or 120 degrees
cannot. The author is convinced that Finding a test signal for measuring
loudspeakers at 150 degrees produce a more envelopment:
exciting surround sound picture. A loud sound
effect from 150 degrees from the front is much This problem turned out to be much more
more exciting than one from 120 degrees, difficult than expected. Much of the difficulty
lies is choosing a test signal that will adequately
In addition, one of the most effective uses of a represent music. As mentioned in the section on
surround sound system for popular music is the concert hall acoustics, the perception of
presentation of a live performance, where the envelopment is highly influenced by the
applause and noise from the audience draws the presence of gaps in the music that allow
listener into the experience. Applause has reverberation to be heard. In this study we will
substantial high frequency content, and the assume that such a gap has already occurred -
front/back ratio becomes important, we are trying to model the transfer function of
the reverberation within that gap.
However, what is good at high frequencies is not
necessarily good at low frequencies. As we will The correlation time of musical signals:
see, for optimal envelopment at low frequencies
the surround loudspeakers should be at the sides However music has another very interesting
of the listening area. We might conclude that a property that is highly relevant to this study.
single pair of surround loudspeakers is not Music generally consists of notes - segments of
sufficient - although if we are clever there may sound that have a recognizable pitch. Although
be ways around this problem, the notes may be rich in harmonics, the
fundamental frequency is often steady.
4. ENVELOPMENT AT FREQIJENCIES
BELOW IO00Hz IN SMALL ROOMS - If we autocorrelate a musical signal with itself
THE DFT we will see that over a certain length of time the
autocorrelation function is strongly non-zero.
Envelopment below 1000Hz depends on That is, as long as a note continues there will be
interaural fluctuations, andon the ability of a strong fluctuation in the autocorrelation
human perception to separate sound into a function. Ando has published material that
found that the average length of the non-zero milliseconds. If the music (or the reverberation
region of the autocorrelation function was related from the music) has a correlation length that is
to the type of music being played, with modern significantly longer than this time constant the
serial music having a short correlation time, and room does not generate interaural fluctuations
romantic symphonies having a long correlation directly. This means that a single driver in the
time. Such a result would be expected from the room will not produce interaural fluctuations at
average length of the notes in the various the ears of a listener. The room can however
musical styles, as well as the presence in many detract from the transfer of fluctuations from
types of modern music of percussive sounds with multiple drivers, and this is the effect we wish to
no well established pitch. Ando discovered that measure.
as the correlation length of a musical example
increased, the ideal reverberation time of a A detector for envelopment at the listening
performance space also increased. We will see position
that this result can be predicted from the
properties of the DFT. A detector for envelopment is also a difficult
problem. We had hoped to be able to use a
When we started working for a measure of measure developed for concert halls, the
envelopment in rooms we used band limited pink interaural difference, or IAD. To make a long
noise as a test signal. Reference [44] contains story short - it doesn't work. We were in fact
several experiments and observations of the unable to find a proxy measure that could even
spatial properties of band limited pink noise, duplicate the perceived envelopment in an
The correlation length of band limited noise anechoic space, let alone the envelopment in a
depends on the bandwidth chosen (and to some reflective space. For example, it is obvious that
degree on the filter type.) For our first work we a single sound source in an anechoic space is
choose to use a bandwidth approximately equal incapable of producing envelopment. Our
to the width of a critical band in the human ear. measure must show that this is the case. By
The results showed that this was not a good similar reasoning, two sources driven by highly
choice. Although the results appeared to be correlated material should also give near-zero
accurately reflect the envelopment of noise envelopment in an anechoic space, and the
signals in small rooms, they did not predict the measure should reflect this. The IAD fails on
envelopment of musical signals, both counts.
As an experiment, we measured the frequency (As luck would have it, the IACC measure does
width and phase fluctuation in the reverberation distinguish between echoic and anechoic spaces.
from musical notes of various lengths in a real However - again making a long story short - the
concert hall (Boston Symphony Hall.) The IACC fails to predict observed results at
resulting reverberation had high coherence. We frequencies below 200Hz.)
were able to use the reverberation in Boston as a
test signal in calculating the DFT, and we found We found it was necessary to go back to first
that to achieve a similar results with noise we principles. The hypothesis in [44] predicts that
had to use a filter bandwidth of 1-2Hz. A the perception of envelopment at low frequencies
disadvantage of such a narrow frequency is that depends on fluctuations in the ITD. There is no
the DFT in a small room is often highly shortcut - to measure envelopment we must
frequency dependent, and to find an average convolve the binaural impulse response from
value of envelopment one must separately each sound source with an independent signal,
calculate the DFT at many different frequencies, and then measure the fluctuations in the lTD that
Unfortunately there appears to be no short-cut, result. In practice this is difficult. Evolution has
We must calculate our measure using a narrow had millions of years to perfect methods of
band test signal, extractinglTD informationfromnoisy signalsat
the eardrums. We had to develop an algorithm
The reason we need a narrow band test signal is in software that make this extraction. Our
obvious in hindsight. Small rooms have current measure still needs some work, but it
relatively short reverberation times, often less seems to give useful results. Using this
than 0.5 seconds. The time constant of such a algorithm and a narrow band noise signal as a
space (the time it takes the sound to decay by test probe we developed our measure for
I/e) is the reverberation time divided by 7, or 70
envelopment in small rooms. We call it the 10. Measure the strength of these fluctuations
DFT, or diffuse field transfer function, by finding the average absolute value of the
fluctuations. The number which results is
The process of finding the diffuse field transfer the Diffuse Field Transfer function, or DFT.
function can be summarized:
11. Measure the DFT as a function of the
1. Calculate (or measure) separate binaural receiver position in the room under test.
impulse responses for each loudspeaker
position to a particular listener position. A
high sample rate must be chosen to maintain The most difficult part of this process is building
timing accuracy. In our experiments the ITD detector in software. The detector must
176400Hz is an adequate sample rate. be robust. The signals at the ears are noise
signals - in many places the amplitude is low,
2. Low-pass filter each impulse response and and the zero crossings can be highly confused.
resample at 11025Hz, and then do it again, Our detector should use very simple elements -
ending with a sample rate of 2756Hz. This just timers and filters - to do the job. It must be
sample rate is adequate for the frequencies very difficult to confuse. The design of this
of interest, and low enough that the detector is beyond the scope of this paper.
convolutions do not take too much time. Persons interested in its design, or in the Matlab
code for the whole DFT measurement apparatus,
3. Create a test signals from independent should contact the author. In the current version
filtered noise signals. Various frequencies of the code, it takes about 15 seconds to find the
and bandwidths can be tried, depending on DFT at a single receiver position, so a 7x7 array
the correlation time of the musical signal of of positions can be calculated in about 12
interest, minutes(ona laptopwitha 150MHzPentium.)
4. Convolve each binaural impulse response The current code uses noise signals with 10000
with a different band filtered noise signal, samples each. This is about 3.7 seconds at the
and sum the resulting convolutions to derive sample rate of 2756Hz. We are interested in
the pressure at each ear. random fluctuations in a signal in the 3Hz to
17Hz band - with typical maximum energy
5. Extract the lTD from the two ear signals by around 5Hz. Needless to say, the accuracy of
comparing the positive zero-crossing time of such a measurement given a 3.5 second duration
eachcycle, is not high. TheDFTweget fa'omthis
measurement has a semi-random variation of
6. Average the ITDs thus extracted to find the about 2dB, with occasional values as many as
running average lTD. The averaging 3dB different from what was expected. Thus the
process weights each lTD by the DFT we present here is somewhat noisy. We
instantaneous pressure amplitude. In other could improve the accuracy by increasing the
words, 1TDs where the amplitudes at the two length of the noise signals - gaining the usual
ears is high count more strongly in the improvement by a factor of 1.4 for each doubling
average than lTDs where the amplitude is of the signal length.
low.
5. CALIBTATION OF THE DFT MODEL
7. Sum the running average lTD and divide by
the length to find the average lTD and the With a single driver in an anechoic space the
apparent azimuth of the sound source. DFT should be zero. In practice, in spite of our
limited length of noise, the values we get are at
8. Subtract the average value from the running least 40dB less than the maximum values with
average lTD to extract the interaural two drivers. Thus our detector passes this test.
fluctuations.
5a. Calibration of the DFT for noise signals
9. Filter the result with a 3Hz to 17Hz
bandpass filter to find the fluctuations that There are two major adjustments to the detector
produce envelopment, that we must set as best as we can to emulate the
properties of human hearing. The most
142 AES15thINTERNATIONAL
CONFERENCE
GENERAL OVERVIEW OF SPATIAL IMPRESSION, ENVELOPMENT, LOCALIZATION, AND EXTERNALIZATION
144 AES15th
INTERNATIONAL
CONFERENCE
GENERAL
OVERVIEW
OFSPATIAL
IMPRESSION,
ENVELOPMENT,
LOCALIZATION,
ANDEXTERNALIZATION
necessary, and then the transforms are added to average pressure is useful nevertheless. To make
find the total pressure and phase. The transform the measure closer to what we hear we weight
of a binaural impulse response is found by the sum by -3dB per octave, so each 1/3 octave
separately calculating the transforms for each band gives an equal contribution to the sum.
ear. At low frequencies we assume the head can
be approximated by two omnidirectional We start with a binaural impulse response, either
receivers, separated laterally by 25cm. We do measured or calculated. That is, a separate
not auempt to model diffraction around the head, impulse response for the left ear and the right ear
which is minimal at the frequencies of interest, of a dummy head.
The lTD as a function of frequency can be found IL(t) = impulse response for the left ear
by finding the phase difference between the IR(t) = impulse response for the right ear
transforms of the left and right binaural Sum = sum of the frequency bins over the
responses. The phase difference is then frequencies of interest
converted to a time difference by dividing by
frequency. 2. FFTL(f) = ffi(IL(t))
3. FFTR(f) = fft(IR(t))
It is useful before going further to examine the
perceptual meaning of the impulse response and 7. Average Pressure ----sqrt(sum((magnitude
its Fourrier transform. Both are mathematical FFTL(0^2)/f)
concepts. It is not obvious that either should
have any relationship to what we hear with This average pressure contains a frequency
music. The impulse response is the response of factor, which we would like to eliminate
the room to a pistol shot or small explosion, when we plot pressure as a function of
These are fortunately rare in everyday life. frequency. We can define a normalized average
Nonetheless there is a considerable body of pressure:
literature that attempts to relate musical
perceptions to direct features of the impulse 8. Normalized Average Pressure = Average
response. Pressure / ( sum (l/sqrt(f))
The transform of the impulse response describes For Iow frequencies we assume that the
the steady state response of the room to single magnitude of FFrL is approximately equal to the
frequency sinusoids. These are plausibly magnitude of FFrR.
musical. In a smallrooms after the time it takes
for the sound to travel across the room a few We suggested above that the ITD at a particular
times the pressure is nearly at the steady state frequency could be found by comparing the
value. Most musical notes are longer than this, phase of the left and right parts of a binaural
and often the attacks of low frequency response. The problem is that this phase
instruments are slow enough that the room can difference is independent of the pressure at that
be modeled by the steady state condition, frequency. It is possible - in fact it is nearly
However each frequency in the transform always the case - that the frequencies where the
represents only a single musical note. We would phase difference is the largest are the ones for
like to know the response of the room over a which the pressure is the lowest. We are not
range of notes. We need a measure for average likely to take much notice of a large lTD if the
pressure and average phase, where the average is pressure is low. Unlike a lateral figure of eight
over frequency, microphone, which gives a maximum output at
the nulls of a lateral standing wave, we cannot
8. NORMALIZED AVERAGE PRESSURE determine the ITD of a sound we cannot hear. If
(NAP), AND AVERAGE lTD (AITD) we stand at such a null, we hear nothing.
To find the average pressure we could simply It would be possible to develop a model that
sum the pressure squared over the range of determines lTD the same way the ear does, by
interest to find the total pressure in the frequency first separating the sound into critical bands, and
band. The summation assumes that each then finding the timing of zero crossings for each
frequency was present in the steady state - which ear. See [44.] As we will see later, such a model
is not particularly likely - but this measure of is quite expensive computationally. Since we
AES15thINTERNATIONAL
CONFERENCE 145
DAVID GRIESINGER
want to calculate the ArrD at many frequencies averages out these differences, giving us a lower
and at many points in a room, there is a strong value than we would get using individual
incentive to develop an efficient model, frequencies. It does reflect the average
localization of a narrow band noise source. We
To make an average lTD that is closer to what are interested in externalization of the sound, and
we actually hear we weight the ITD with the externalization can take place even in the
pressure, lTDs with Iow pressure are given low absence of correct localization. Thus for
weighting, and ITDs with high pressure are externalization the average absolute value of the
weighted more strongly. We then normalize the ITD, the A1TD is a better measure.
resulting sum by dividing by the average
pressure ox)er the same frequency range. At low frequencies in a free field - with no
reflections - AITD and NliD are equal, and
Given this description the method of finding the independent of frequency. Their value depends
AITD follows directly: on the angle between the listener and the source,
and can be easily calculated.
5. Phase_angle(f) = angle(FFTL)-angle(FFTR)
6. liD(f) = Phase_angle(f)/(2*pi*f) If source_angle is the angle between the source
and a line drawn between the listener's ears
To find the AITD we multiply the absolute value (listener facing forward):
of lTD(f) by the absolute value of the pressure,
and sum over frequency. We also apply a -3dB Lateral AITD = 0.75ms*cos(source_angle)
per octaveweighting to give equal weight to all
bands. The sum is then normalized by dividing
by the weighted sum of the absolute pressure
overthe samefrequencyrange, os
6. AITD= ._°'s1
sum(magnitude(FFTL(f)) *abs(ITD(f))/(sqrt( <2o.4.
f))/(average pressure) _..J 02,
It is also possible to define a NET lTD (NITD) - 0,
one which preserves the sign of the ITD in the 2.54.5 6.5 5.5
sum. The NITD is always less than the AITD, 8.510.5
12.5
but it indicates our ability to determine the lateral
position-
ft 14.5 17.5
direction of a sound source, not just its position-ft
front-back
externalization.
Figure 1: The Lateral AITD in the center of a
7. NITD = 17'x23' anechoic space with a driver in the
sum(magnitude(FFTL(f))*ITD(f)/(sqrt(f))/(a upper left comer (at. 1',. 1'). Where the value is
verage pressure) high, the sound source will be localized to the
side of the listener.
Note that for simplicity we use only the
magnitude of FFTL(f) to represent the pressure We can also define a Medial AITD - by simply
at both ears. We could average the magnitudes rotating the listener's head by 90 degrees in
of FFTL and FFTR, but there seems to be no azimuth and calculating the AITD again. In a
reason to do so in practice, free field - assumingthe source angle has not
changed:
9. AITD AND NITD - WHAT DO THEY
MEAN?
Acoustics. Proc. 8th AES Conference Wash. DC 22. Griesinger, D., Room Impression,
May 1990 pp 59-69 Reverberance, and Warmth in Rooms and Halls.
12. Bradley, J.S., Contemporary approaches Presented at the 93rd Audio Eng. Soc.
to evaluating Auditorium Acoustics. Proceedings convention in San Francisco, Nov. 1992. AES
of the Sabine Conference, MIT June 1994. preprint #3383
13. Bradley, J.S. and Soulodre G.A., 23. Griesinger, D., Progress in
Spaciousness judgments of binaurally electronically variable acoustics. Proceedings of
reproduced sound fields. Ibid. p 1-13 the Sabine Conference, MIT June 1994.
14. Bradley, J.S., Comparisons of IACC 24. Griesinger, D., Subjective loudness of
and LF Measurements in Halls. 125th meeting of running reverberation in halls and stages.
the Acoustical Society'of America, Ottawa, Proceedings of the Sabine Conference, MIT June
Canada,May1993 1994.p89
15. Bradley, J.S., Pilot Study of Simulated 25. Griesinger, D., Quantifying Musical
Spaciousness. Meeting of the Acoustical Society Acoustics through Audibility. Knudsen
of America, May 1993 Memorial Lecture, Denver ASA meeting, 1993.
16. Bradley, J.S. and Souloudre, G.A., 26. Griesinger, D., Further investigation
Objective measures of Listener Envelopment. J. into the loudness of running reverberation.
Acoust Soc. Am. 98 [1995] pp 2590-2597 Proceedings of the Institute of Acoustics (UK)
17. Bradley, J.S. and Souloudre, G.A., conference, Feb 10-12 1995.
Listener envelopment: An essential part of good 27. Griesinger, D., How loud is my
concert hall acoustics. JASA 99 [1996] p 22 reverberation. Audio Engineering Conference,
18. Gardner, W. and Griesinger, D., Paris, March 1995. Preprint # 3943
Reverberation Level Matching Experiments' 28. Griesinger, D., Design and performance
Proceedings of the Sabine Conference, MIT June of multichannel time variant reverberation
1994. p263 enhancement systems. Proceedings of the Active
19. Gold, M.A., Subjective evaluation of 95 Conference, Newport Beach CA, June 1995
spatial impression: the importance of 29. Griesinger, D., 'Optimum reverberant
lateralization, ibid. p97 level in halls' proceedings of the International
20. Griesinger, D. Measures of Spatial Congress on Acoustics, Trondheim, Norway
Impression and Reverberance based on the June 1995
Physiology of Human Hearing. Proceedings of 30. Hartman, W., Localization of a sound
the 1lth International AES Conference May source in a room. Proceedings of the 8th
1992 p114 - 145 international conference of the Audio
21. Griesinger, D., IALF - Binaural Engineering Society, May 3-6 1990
Measures of Spatial Impression and Running 31. Hidaka, T. and Beranek, L. and Okano,
Reverberance. Presented at the 92nd Convention T., Interaural cross correlation (IACC), lateral
of the AES March 1992 Preprint #3292
fraction (LF), and sound energy level (G) as 35. Kreiger, A., Nachhallzeitverlangerung
partial measures of acoustical quality in concert in der Deutschen Staatsoper Berlin. Tonmeister
halls. J. Acoust Soc. Am. 98 [1995] pp 988-1007 Informationen (TMI) Heft 3/4 Marz April 1996.
32. Jullien, J.P. and Kahle, E. and 36. Morimoto, M. and Maekawa, Z.,
Winsberg, S. and Warusfel, O., Some Results on Effects of Low Frequency Components on
the Objective Characterization of Room Auditory Spaciousness Acustica 66 [1988] pp
Acoustical Quality in Both Laboratory and Real 190-196
Environments Proc. Inst. of Acoustics Vol. 14 37. Morimoto, M. and Posselt, C.,
part 2 (1992). presented at the Institute of Contribution of Reverberation to Auditory
Acoustics conference in Birmingham, England, Spaciousness in Concert Halls. J. Acoust. Soc.
May1992 Jpn(E)10[1989] 2
33. Kahle, E., Validation d'un modele 38. Morimoto, M. and Maekawa, Z.,
objectif de la perceptionde la qualite acoustique Auditory Spaciousness and Envelopment. 13th
dans un ensemble de salles de concerts et ICA, Yugoslavia 1989
d'operas' Doctorate Thesis, IRCAM June 1995 39. Morimoto, M., The relation between
34. deV. Keet, W., The Influence of Early spatial and cross-correlation measures 15th ICA
Lateral Reflections on the Spatial Impression. Norway 1995
The 6th International Congress on Acoustics, 40. Schroeder, M.R. and Gottlieb, D. and
Tokyo, Japan, Aug 21-28 1968. pp E53 to E56 Siebrasse, K.F., Comparative study of European
AES
15th
INTERNATIONAL
CONFERENCE 149